You cannot get the newer version of those rows inside Sync Framework. as far as its concerned, the rows that were inserted were the rows that it has retrieved from the client. it has no idea of the manipulation you did in the dataset to remove the original
PK column.
you will have to query the server database directly. your problem though is how to retrieve the same rows you have in the dataset from the server since you modified the PK already. the originator id simply gives you which client a row has come from. it doesnt
give you the specific rows within the sync session.
had you inserted the PK value (in a non-PK column in the server) + originator id, you could use the combo to query the server to retrieve the new PK.